Import Qemu/KVM VMs from Ceph/RBD backend

[wilschie]

Hi,

I have many Qemu/KVM virtual machines running on non-PVE servers with virtual disks on an external shared Ceph RBD backend. I want to move these machines to a PVE infrastructure. The PVE servers will have access to the Ceph storage backend but should not manage it. Is it possible to migrate the existing virtual machines with their RBD images to the PVE servers without lengthy import processes (qm importdisk), perhaps simply by renaming the disks and editing VM configuration files?

Thank you very much.

Best regards
Willi

 

[Uplink]

Are you using Ceph namespaces? IIRC namespaces are meant to separate storage access for disparate environments. If you are, you'll need to map the namespace the VMs are in to Proxmox.

PVE doesn't need to manage Ceph in order to use it. That shouldn't be an issue.

If you can rename the disks in Ceph, and if you modify the vmid.conf files, in theory that might work. I'm not sure how Proxmox hands Ceph on the backend - though I've done something similar with NFS mounts.

The safest option really is doing a back-up of all the VMs, and then importing them. There are built-in features for this, and it can be easily scripted. This way you're guaranteed everything will work in a standardized process